Introduction

Marie Toft founded Emotionise AI in December 2018, drawing on her background as a TV producer and documentary maker for BBC, SBS Australia, and RTE. Despite her successful media career, Marie pursued further education in psychotherapy, where she discovered the significance of emotional connections in communication.

In 2020, she combined her skills in communication and psychotherapy with the potential of Generative AI, aiming to train AI in emotional intelligence. This initiative led to the creation of Emotionise AI. In 2022, experienced tech founder Tim Pat Dufficy and Emotionise CTO Siva Sharma joined Marie as co-founders.

Since their collaboration, Emotionise AI has secured over €500k in investments and developed AI capable of mimicking emotional intelligence. Their focus on enhancing Employee Engagement and Customer Support has yielded positive outcomes. Emotionise's partnership with Allion since January 2023 has been instrumental in their success.

Generative AI

Marie's recognition of Generative AI's potential in June 2020, with the launch of GPT-3 by OpenAI, sparked her vision for Emotionise AI. She understood the opportunity to leverage Large Language Models (LLMs) to imbue AI with emotional intelligence. During the development of their MVP in 2021 and 2022, Marie took charge of creating the emotionally intelligent dataset needed to train the model.

This process shed light on the critical role of quality data in effectively training LLMs, prompting Emotionise to assemble a team of writers with strong communication skills and emotional intelligence. Emotionise's foresight in harnessing Generative AI predates the widespread recognition catalyzed by the release of ChatGPT, positioning them as early pioneers in the field.

The Goal

The aim was ambitious, to craft the world's inaugural AI and Natural Language Processing (NLP) domain model infused with emotional intelligence as its foundation.

In January 2023, collaborating closely with the accomplished team at Allion, Emotionise embarked on a mission. Their objective was clear, to train AI to dynamically and adeptly respond with emotional intelligence to a spectrum of communication, spanning both written and verbal interactions. This endeavor represented a significant leap forward in the realm of AI, promising to reshape how technology engages with and understands human emotions.

Our Methodology

Before commencing any project at Allion, our team initiates by crafting a conceptual diagram, which serves as a visual depiction of ideas and interrelationships within the project. This diagram aids us in mapping out crucial components and processes, ensuring a comprehensive understanding. By incorporating our main concept into the diagram, we attain a clear overview of the project's theme. This visual aid plays a pivotal role in elucidating how different components intertwine and contribute to achieving the project's objectives. Furthermore, it enables us to establish connections between various aspects of the project, such as the user journey and model performance, thereby facilitating enhanced decision-making and communication among team members.

Implementation

Having acquired a thorough understanding of the project, our team defines the development approach. This process involves grasping the project's objectives, requirements, and stakeholders' needs, as well as deciding on methodologies, technologies, and strategies for development. To manage complexity, we divided the application into four parts:

1. Web application development: This includes building the user interface and client-side functionality.
2. Plugin development: Involves creating modular components to enhance functionality.
3. Backend development: Focuses on implementing server-side logic, database interactions, and security measures.
4. Ensuring secure data processing: Involves implementing encryption, access controls, and authentication mechanisms.

Security was integrated into the design phase to proactively address potential vulnerabilities and implement best practices.

In terms of tools and technologies:

• Version Control System: Bitbucket was used for managing code repositories, version control, and collaboration.
• Frontend Development: JavaScript was the chosen programming language, with React.js as the framework/library. Additional tools included Redux for state management, Axios for making HTTP requests, and Material-UI or Bootstrap for building responsive user interfaces.
• Backend Development: C# was used as the programming language for .NET Core, with additional tools such as Entity Framework Core for ORM and ASP.NET Core for building web APIs and server-side logic.
• Database: MySQL Server was utilized as the RDBMS for storing and managing data.
• Plugin Development: React.js was used for frontend plugins, with additional tools depending on requirements.
• Security: Authentication and authorization were handled using JSON Web Tokens (JWT) and OAuth 2.0, while SSL/TLS was used for secure communication. Additional security testing was conducted using OWASP ZAP and Microsoft Identity Platform for identity management in Azure.

Development was conducted using Visual Studio or Visual Studio Code as the IDE, with Node Package Manager (npm) for managing frontend dependencies and NuGet for managing .NET dependencies. Build tools included Webpack or Parcel for bundling frontend assets and MSBuild for building .NET Core projects.

For continuous integration/continuous deployment (CI/CD), GitHub Actions was used to automate build, test, and deployment processes. Monitoring and logging were facilitated using Azure Application Insights for monitoring application performance and Serilog for logging events and exceptions. Testing procedures were employed based on discussed methodologies, with measures taken to ensure solution quality.

User Feedback

Between October 12th and November 3rd, 2023, Emotionise conducted a pilot beta testing program of Emotionise AI tool within the Irish branch of a global pharmaceutical company in collaboration with The Outcome, its research partner. Multiple users from various departments participated, including manufacturing, HR, and senior leadership. They engaged in pre-testing interviews, feedback surveys, check-in calls, and final feedback interviews.

Initial findings indicate emotional benefits reported by testers, the tool was most effective at supporting them in crafting an authentic message, making the message more engaging, increasing the emotional intelligence of the message, and increasing the clarity and conciseness of the message. One tester highlighted the tool's opportunity for continued learning, stating its potential to improve communication, emotional intelligence, and social skills through personalized suggestions and explanations.

Testimonial

“Emotionise has been working closely with Allion since Jan 2023. As our technical partner, they are a vital part of our team and success. They are also investors in our company. Working alongside them, we have built the world’s first and only AI & NLP domain model trained with emotional intelligence at its core. We create bespoke models trained on an organization’s values to improve outcomes for Employee Engagement and Customer Support. But there is no vertical that increased emotional intelligence will not improve and the potential and scale for our AI is huge.

We have beta tested our model and the results exceeded our expectations. 2024 is the year of further paying customers and metrics and we are in discussions with companies such as Pfizer and Amazon. literally could not have achieved what we have without Allion. We have the benefit of an incredibly dedicated and talented technology team. We cannot recommend Allion Technologies highly enough.” – Marie Toft, CEO & Co-Founder Emotionise AI.